This invention relates generally to amplification systems and more particularly to methods and apparatus for reducing distortion in amplifiers used in such systems.
As is known in the art, amplifiers have a wide variety of applications. Amplifiers can be biased to operate in one of a number of so-called Classes. When biased to operate in Class A, the amplifier provides a linear relationship between input voltage and output voltage. While operation in Class A has a wide range of applications, when higher power output and efficiency are required or desired, the amplifier is sometimes biased to operate in Class A/B. When biased to operate in Class A/B, however, the Class A/B amplifier power transfer curve 10 is less linear than for Class A amplifiers, illustrated in FIG. 1 by trace 14. To increase efficiency, communication systems often operate amplifiers in the non-linear region 12. This practice, however, does introduce amplitude and phase distortion components into the output signal produced by the amplifier.
As is also known in the art, most communication systems have FCC allocated frequency bandwidths 18 (that is, in-band frequencies) centered about a carrier frequency 20 as shown in FIG. 2A. For example, a CDMA (Code Division Multiple Access) communication system signal has a predefined bandwidth of 1.25 MHz. Different CDMA communication channels are allocated different bands of the frequency spectrum. Amplifiers are used in such systems, and are frequently biased to operate in Class A/B. Referring to FIG. 2B, signal processing such as amplification by an amplifier operating in the non-linear region 12 (FIG. 1) can produce distortion frequency "shoulders" 22a-22b outside a signal's allocated bandwidth 18. (These are called out-of-band frequencies.) These distortion frequency components 22a-22b can interfere with bandwidths allocated to other communication signals. Thus, the FCC imposes strict limitations on out-of-band frequency components.
Many techniques exist to reduce out-of-band distortion. One such technique is shown in FIG. 3 where a predistortion unit 24 is fed by a signal 25 to be amplified. The predistortion unit 24 has a power transfer characteristic 24a (FIG. 1) and compensates for distortion introduced by subsequent amplification in Class A/B amplifier 26. More particularly, the predistortion unit 24 transforms electrical characteristics (for example, gain and phase) of the input signal such that subsequent amplification provides linear amplification to the phase and frequency characteristics of the input signal. The predistortion unit 24 is configured with a priori measurements of the non-linear characteristics of the Class A/B amplifier. Unfortunately, the amplifier characteristics (amplification curve 10 with region 12 of FIG. 1) change over time and temperature making effective predistortion more difficult. For example, as the temperature of the amplifier increases, its non-linear region 12 may become more or less linear, requiring a compensating change in the transform performed by a predistortion unit 24. Some adaptive predistortion systems use look-up tables to alter predistorter characteristics based on environmental factors such as temperature. These look-up tables include predetermined predistorter control settings for use in predetermined situations. However, environmental factors alone do not determine the alterations in an amplifier's characteristics. Thus, over time, amplifier characteristics vary unpredictably due to aging of amplifier components.
Another approach to reduce amplifier distortion is to use feedforward compensation, as shown in FIG. 4. Here, a feedforward network 31 is included for reducing out-of-band distortion. The feedforward network 31 includes a differencing network or combiner 30, a main amplifier 33 operating as a Class A/B amplifier, an error amplifier 32, delay circuits 28 and 28a, and a combiner 29. The differencing network 30 produces an output signal representative of the difference between a portion of the signal fed to the amplifier 33 operated Class A/B and the signal fed to the amplifier 33 prior to such amplification. The frequency components in the differencing network 31 output signal are, therefore, the out-of-band frequency components 22a-22b introduced by amplifier 33. Amplifying and inverting the output produced by the differencing network 30, by error amplifier 32, produces an out-of-band correcting signal. More particularly, the combiner 29 combines the correcting signal produced by differencing network 30 and amplifier 32, with the delayed signal output of amplifier 31 thus reducing the energy in the out-of-band frequencies 22a-22b (FIG. 2B) of the signal output by amplifier 33. Feedforward network 31 includes delay line 28 to compensate for the delay in error amplifier 32. It should be noted that minute differences in timing between these elements can impair the effectiveness of a feedforward system. While a manufacturer can carefully match components prior to shipment, as feedforward components age, the correcting signal and processed signal can become mistimed if not properly compensated.